SDFormat in the ROS ecosystem

[azeey]

Users in the community need guidance on how best to use URDF with SDFormat, especially when using URDF with tools in the ROS ecosystem together with Gazebo. This issue is meant to capture some of the questions/requests raised by members of the community and as a place for discussion on how to address them. There is also an ongoing effort to improve the interoperability of SDFormat with tools in the ROS ecosystem. Since there's likely to be overlap between the work there and some of the discussion here, this issue will hopefully serve as a central place for all discussions regarding SDFormat in the ROS ecosystem.

Here's an initial list of questions to address/discuss:

• What are some best practices for mapping SDFormat models (with nested models) to URDF?
  • Is this even the best way to go about this? Should tools like RViz work directly with SDFormat?
• Can we create a tool to replace model:// to package://?
• Do we have examples showing how to use URDF with ROS and the new Gazebo? If so, how can we improve their visibility?
• Would it be possible to only use SDFormat (and not URDF) with Gazebo and the ROS ecosystem?

cc @srmainwaring @quarkytale @bperseghetti

[peci1]

The CTU-CRAS-Norlab models in SubT use a fully automated pipeline to "render" various outputs (URDF for rviz, SDF for G classic or for new G) from a single Xacro source. The pipeline is probably not easy to understand for newcomers, though...

See e.g. https://github.com/osrf/subt/tree/master/submitted_models/ctu_cras_norlab_marv_sensor_config_1 .